Often a Pandas-build HDF5 file has a sorted column that we can use to cleanly and efficiently partition the dataset. We should expose this case through the read_hdf function.

cc @jreback @jcrist